Run coding agents without babysitting them. Keep jobs focused, bounded, checked and accountable from start to finish. Finally run your agents swarms overnight and get your time back.
-
Updated
Oct 19, 2026 - TypeScript
Run coding agents without babysitting them. Keep jobs focused, bounded, checked and accountable from start to finish. Finally run your agents swarms overnight and get your time back.
Turns repeatable, domain-agnostic workflows into graph-driven loops.
Runtime evidence layer for agentic AI — signed audit chain, deterministic gate verdicts with record-bound audit replay. pip install phionyx-core.
Agent Gate Incident Replay is a browser-runnable blackbox replay runtime for Agent incidents. It restores a real VM state in the browser with [v86](https://github.com/copy/v86), mounts an incident module, and replays the agent's actions against real runtime boundaries and real Agent Gate verdicts.
Offline verifiers for the audit evidence of Zarel, the governed runtime for AI agents that act on money and regulated records: hash chain, signed checkpoints, RFC 3161 anchoring, RFC 4998 evidence records.
The TypeScript SDK for Zarel, the governed runtime for AI agents that act on money and regulated records: one client with runtime and contract namespaces, zero runtime dependencies.
Hamiltonian Programming Language: scheduler-governed execution, typed refusal, deterministic evidence and cryptographic anchoring.
The zarel CLI for Zarel, the governed runtime for AI agents that act on money and regulated records, and the offline verifier for its audit evidence.
To associate your repository with the governed-runtime topic, visit your repo's landing page and select "manage topics."